Hello, I'm not a mathematician, so take my answers with a grain of salt:
1. You should certainly have your FIR bin length equal to your TR. With
longer bins, you are trying to match two time points to a single value.
This would be fine if your signal were constant, but in highly variable
regions (such as the rising or falling phases of an HRF), this would
most likely result in SPM fitting the mean of those two time points. The
difference between the two time points and that mean would be included
in the residual and therefore increase your error.
2. I'm not sure; it depends on if my answer to 3 is right. If you can
use those peak points, an ANOVA on those points seems like the obvious
answer.
3. Yes, I think that if fine.
--
Kevin T. Hill
Ph.D. Candidate
Lee Miller's Lab
Center for Mind and Brain
University of California, Davis
(530) 297-4426
http://mindbrain.ucdavis.edu/content/Labs/Miller/
|